Yann LeCun, Facebook’s chief AI scientist and an early machine-learning architect, said the expanded research effort was pushed by Facebook leaders such as CEO Mark Zuckerberg. “AI has become so central to the operations of companies like ours, that what our leadership has been telling us is: ‘Go faster. You’re not going fast enough,’ ” LeCun said.
Facebook’s international expansion and academic partnerships mimic the model used by many of AI’s biggest corporate players, which gives top researchers and engineers the freedom to teach classes, publish papers and split their time between academic and commercial pursuits.
LeCun said Facebook’s AI researchers will focus on the long-term pursuit of “machines that have some level of common sense” and learn “how the world works by observation, like young children do in the first few months of life.”
AI — the sweeping term for systems that train on lots of data, make decisions and improve without overt human control — is one of modern technology’s most competitive fields, with tech giants such as Google, Amazon and Microsoft grappling with firms in China and Europe for the top talent, applications and ideas. Google CEO Sundar Pichai in January called AI “one of the most important things that humanity is working on” and “more profound than electricity or fire.”
While Facebook’s AI rivals have developed automated voice assistants, such as Amazon’s Alexa, or spearheaded breakthroughs in gaming, robotics or self-driving cars, the social network has devoted much of its AI engineering to analysis of images, video and text — facial recognition, language translation and detection of unwanted content or harmful comments. (Amazon.com founder and chief executive Jeffrey P. Bezos owns The Washington Post.)
A typical user can often miss the subtle ways Facebook’s AI influences their actions, including photo tag suggestions, algorithmically decided news feeds, friend recommendations, spam blocking and ad targeting.
AI is a critical battlefield for Facebook, with Zuckerberg pledging to Congress and investors that automated tools would help solve some of the company’s thorniest problems, including extremist propaganda, misinformation and hate speech. The company says AI has boosted its ability to monitor the social network’s 2 billion users, but it still relies heavily on human moderators and, as LeCun noted, “works not so well with false news.”
The company on Tuesday announced new partnerships in Pittsburgh, Seattle, London and Facebook’s hometown, Menlo Park, Calif., to bring on specialists in developing AI to scour vast image libraries, process varying languages and analyze human motion.
Facebook’s global expansion, beginning with labs in Paris and Montreal three years ago, has allowed the company to secure a seat near top universities, research institutes and other “talent factories” where AI researchers and academics are at work.
Google recently announced it would open an AI lab in Accra, the capital of Ghana, in hopes of attracting AI talent from across West Africa.
Details of Facebook’s AI research are limited. But LeCun said one lab — in Pittsburgh, known for its Silicon Valley-like “Robotics Row” — will study robots with the goal of further understanding the interplay between humans and machines.
The research, LeCun said, could be used to build better virtual assistants, though he said it was unlikely anyone would be chatting with a Facebook robot anytime soon.
“It’s too early to talk about consumer applications of robotics, but robots will likely play a role in bringing people together,” he said.